Resolve Persistent HTTP 429 (Too Many Requests) by Re-authenticating OAuth Session

[alieonsido]

TL;DR:
alias gemini="if [ -f ~/.gemini/oauth_creds.json ]; then jq 'del(.access_token) | .expiry_date = 0' ~/.gemini/oauth_creds.json > ~/.gemini/oauth_creds.tmp && mv ~/.gemini/oauth_creds.tmp ~/.gemini/oauth_creds.json; fi; \gemini"

Put it to your .bashrc or any CLI Settings to let gemini server force to refresh access_token when you execute gemini in CLI.
It will mitigate the 429 problem at least half an hour in my situation.

Problem Overview

• Widespread Issue: Many users are encountering frequent HTTP 429 (Too Many Requests) errors when using the gemini-3.1-pro-preview model in gemini-cli.

• Unsustainability of Downgrading: While a common suggestion in the community is to downgrade the version, this is not a sustainable solution for long-term development.

• Effective Workaround: Testing has shown that re-authenticating (re-logging via OAuth) can immediately and effectively resolve this rate-limiting state.

Technical Hypothesis

• Credential Decay: This phenomenon may not be actual traffic congestion. Instead, expired or corrupted OAuth tokens might cause the server to relegate requests to a default pool with extremely low quotas.

• Quota Re-association: Re-running the login process to obtain a fresh token forces the server to re-associate the requests with the correct Google Cloud project quota.

• Cache Clearance: Re-authentication also clears local cache issues that might lead to background retry loops, which often trigger automatic API gateway protection.

Suggested Workaround for Users

If you are facing persistent HTTP 429 errors, please try the following before considering a downgrade:

1. Log out of your current session or clear existing credentials.

2. Re-run the OAuth login process (e.g., executing the auth login command).

3. Ensure that your default project is correctly configured in your environment variables or config file.

Result: This process typically PERHAPS assigns a fresh request counter and session quota, restoring normal operation immediately.

Request for Feedback

I encourage other users facing this issue to report if this workaround works for them. This will help the maintainers determine if the root cause lies in credential management or the backend API's quota allocation logic.

[gemini-cli]

Found possible duplicate issues:

• Persistent Status 429s Too Many Requests for last 2 days #23900

If you believe this is not a duplicate, please remove the status/possible-duplicate label.

[arlucio]

Didin't worked for me, I got Ultra subscription and have been 2 weeks unable to use any model via gemini-cli, only getting 429

[maxspevack]

 ▝▜▄     Gemini CLI v0.36.0-preview.7
   ▝▜▄
  ▗▟▀    Signed in with Google: REDACTED_EMAIL /auth
 ▝▀      Plan: Gemini Code Assist in Google One AI Ultra

As recommended in #24384, running /auth signout and /auth signin solved the issue for me. I would rate the response time on gemini-3.1-pro-preview as "barely acceptable" but certainly better than the unusable minutes and minutes of nothing from the last few days.

EDIT: Spoke way too soon. Fix only worked very briefly. Now attempting #24396 but also without success. What a mess.

[jjiangweilan]

it seems this solved the issue for me. I have tried deleting the auth json under .gemini folder, but that didn't work. /auth signout and /auth signin worked

[alieonsido]

Didin't worked for me, I got Ultra subscription and have been 2 weeks unable to use any model via gemini-cli, only getting 429

Do you mind to talk about that in details?
The method I mentioned is to clean cache & re-order the request queue in my guess.
So if you did /auth logout & /auth login I guess it'll be better than you don't do anything?

[alphaply]

I canceled my subscription and subscribed to Copilot instead.😂

/auth to re-authenticate doesn't work for me.

[mehtapd]

It did not work for me.
2.5 and 3-flash work.
3.1-pro-preview does not work.
I have Ultra subscription.

[HPSDf]

It did not work for me. 2.5 and 3-flash work. 3.1-pro-preview does not work. I have Ultra subscription.

same

[alieonsido]

I just retried Gemini-Cli and had experienced work normally and HTTP 429, both of them.
But, when I was experiencing 429 problem, I tried to re-login again, and it works for a while. (Remember: /auth signout and /auth signin should not be the same gemini application, log out and log in are separated perhaps raise the success rate)
so, I guess when you stuck, all you thing you need to do is re-login so that you'll get a pass in a mean time, but the time is not stable. It could be short for minutes or longer than half an hour in my situation.
Work with /resume command to restore chat & history commands shall be a temporarily solution now.

[hardlyh]

After testing, this solution works, but how long it will last is uncertain.

[tuapuikia]

After testing, this solution works, but how long it will last is uncertain.

until your next limit reset. This trick won't work, the rate limit is done in the backend and you can't spoof or reset usage metrics from the google server.